Federated learning is a distributed artificial intelligence framework, which allows multiple edge devices to train a single model collaboratively. In this talk, we first introduce a personalized federated learning algorithm which can tackle the issues of data heterogeneity and device heterogeneity. Then, we present a content-based viewport prediction framework for 360-degree video streaming, wherein users’ head movement prediction models are trained using a personalized federated learning algorithm. The output of the viewport prediction framework corresponds to which video tiles to be transmitted. Finally, we present an algorithm to determine the bitrate and beamforming matrices in a THz-enabled 360-degree video streaming system with multiple access points.
